roberta-large-semeval2012-average-prompt-e-nce-classification-conceptnet-validated

relbert/roberta-large-semeval2012-average-prompt-e-nce-classification-conceptnet-validated

RelBERT fine-tuned from roberta-large on
relbert/semeval2012_relational_similarity. Fine-tuning is done via RelBERT library (see the repository for more detail). It achieves the following results on the relation understanding tasks:

• Analogy Question (dataset, full result):
  • Accuracy on SAT (full): 0.5213903743315508
  • Accuracy on SAT: 0.5222551928783383
  • Accuracy on BATS: 0.6292384658143413
  • Accuracy on U2: 0.4649122807017544
  • Accuracy on U4: 0.5277777777777778
  • Accuracy on Google: 0.768
• Lexical Relation Classification (dataset, full result):
  • Micro F1 score on BLESS: 0.9121591080307367
  • Micro F1 score on CogALexV: 0.8328638497652581
  • Micro F1 score on EVALution: 0.652762730227519
  • Micro F1 score on K&H+N: 0.9641093413090353
  • Micro F1 score on ROOT09: 0.8827953619554998
• Relation Mapping (dataset, full result):
  • Accuracy on Relation Mapping: 0.8159126984126984

Usage

This model can be used through the relbert library. Install the library via pip

pip install relbert

and activate model as below.

from relbert import RelBERT
model = RelBERT("relbert/roberta-large-semeval2012-average-prompt-e-nce-classification-conceptnet-validated")
vector = model.get_embedding(['Tokyo', 'Japan'])  # shape of (1024, )

Training hyperparameters

The following hyperparameters were used during training:

• model: roberta-large
• max_length: 64
• mode: average
• data: relbert/semeval2012_relational_similarity
• split: train
• data_eval: relbert/conceptnet_high_confidence
• split_eval: full
• template_mode: manual
• template: I wasn’t aware of this relationship, but I just read in the encyclopedia that <obj> is <subj>’s <mask>
• loss_function: nce_logout
• classification_loss: True
• temperature_nce_constant: 0.05
• temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'}
• epoch: 30
• batch: 128
• lr: 5e-06
• lr_decay: False
• lr_warmup: 1
• weight_decay: 0
• random_seed: 0
• exclude_relation: None
• exclude_relation_eval: None
• n_sample: 640
• gradient_accumulation: 8

The full configuration can be found at fine-tuning parameter file.

Reference

If you use any resource from RelBERT, please consider to cite our paper.

@inproceedings{ushio-etal-2021-distilling-relation-embeddings,
    title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels",
    author = "Ushio, Asahi  and
      Schockaert, Steven  and
      Camacho-Collados, Jose",
    booktitle = "EMNLP 2021",
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
}